I believe that Chromium is using Skia's reported font metrics which are
currently calculated at
https://skia.googlesource.com/skia/+/00dcf668659ab26b8691c64a977fdd19e9e8df09/src/ports/SkFontHost_FreeType.cpp#1411
. The biggest issue is that historically (and as noted above) FT_Face would
always use HHEA metrics if they weren't 0, and then fallback to usWin or
typo and UseTypoMetrics wasn't looked at to change the priority. As a
result Skia has code to look at the UseTypoMetrics bit and if set reads the
values from the (FreeType generated) OS/2 table if available instead of
using the FT_Face values. So if FreeType interpolates the TT_OS2.sTypo*
values in the table it creates for an FT_Face then that should get picked
up by any font which sets UseTypoMetrics. Any font which doesn't set
UseTypoMetrics will get whatever the FT_Face says to do (except bitmap only
fonts use FT_Face.size.metrics values).

I think it would be great for FreeType to look at the UseTypoMetrics bit.
This would allow for the (eventual) simplification of the Skia code.

Le mer. 30 janv. 2019 à 20:21, Nikolaus Waxweiler <[email protected]> a
écrit :

> The short version is that Firefox and Chromium seem to directly access
> hhea and typo themselves. The patches seem to have no effect except that
> Chromium seems to react to freetype not blindly setting typo values in VF
> instances anymore, but only with small metrics changes.
>
> GTK and Qt let text and UI elements compress if bit 7 is set and typo
> ascender plus descender < hhea ascender plus descender. I have no idea how
> much this is the result of theming in either toolkit and if it is what is
> supposed to happen.
>
> So.
>
> Still not sure we want mvar typo deltas to modify currently active metrics?
>
> ... Merge and break out asbestos pants?
> _______________________________________________
> Freetype-devel mailing list
> [email protected]
> https://lists.nongnu.org/mailman/listinfo/freetype-devel
>
_______________________________________________
Freetype-devel mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/freetype-devel

Reply via email to